There has been a marked change in the way I translate in the past year. I have now become a confirmed reviser or « post-editor » of machine translation (MT) rather than translating from scratch. This has happened since machine translation was integrated into the workflow of the translation environment tool that I use, Déjà Vu.
It is simply faster to revise a machine translation today than to do all the work myself. This was definitely not the case when I tried the same experiment some ten or fifteen years ago. I started using MT in May 2011.
Not only is MT post-editing faster but there is a quality improvement as well. I am impressed by the idiomatic quality of many of the phrases proposed by machine translation systems. Often the suggested translations of parts of sentences are more natural in the target language than what I may have thought of myself. And for technical terms, it frequently points me in the right direction by suggesting translations I am not familiar with. I need to do further research to confirm the suggestions, but having a good starting point for research saves time and improves quality.
Of course on the level of sentences, the translation is often inadequate or downright unintelligible. It reminds me of a t-shirt I once saw « I’m not perfect, but parts of me are excellent. » I build on the good parts to put together an edited translation that takes less time and is better quality than what I could do without the machine’s help.
This is my experience of machine translation post-editing on a practical level using current technology. Machine translation represents an excellent assistant and cost-effective tool for an experienced translator.
